The Astronomical Journal, 137:4824–4833, 2009 June
C 2009.
doi:10.1088/0004-6256/137/6/4824
The American Astronomical Society. All rights reserved. Printed in the U.S.A.
THE DISCOVERY OF A MASSIVE CLUSTER OF RED SUPERGIANTS WITH GLIMPSE
Michael J. Alexander1 , Henry A. Kobulnicky1 , Dan P. Clemens2 , Katherine Jameson2 , April Pinnick2 ,
and Michael Pavel2
2
1 Department of Physics and Astronomy, University of Wyoming, Laramie, WY 82071, USA
Institute for Astrophysical Research, 725 Commonwealth Avenue, Boston University, Boston, MA 02215, USA
Received 2008 November 6; accepted 2009 March 12; published 2009 April 29
ABSTRACT
We report the discovery of a previously unknown massive Galactic star cluster at = 29.◦ 22, b = −0.◦ 20. Identified visually in mid-IR images from the Spitzer GLIMPSE survey, the cluster contains at least eight late-type
supergiants, based on follow-up near-IR spectroscopy, and an additional 3–6 candidate supergiant members having
IR photometry consistent with a similar distance and reddening. The cluster lies at a local minimum in the 13 CO
column density and 8 μm emission. We interpret this feature as a hole carved by the energetic winds of the evolving
massive stars. The 13 CO hole seen in molecular maps at VLSR ∼ 95 km s−1 corresponds to near/far kinematic
distances of 6.1/8.7 ± 1 kpc. We calculate a mean spectrophotometric distance of 7.0+3.7
−2.4 kpc, broadly consistent
with the kinematic distances inferred. This location places it near the northern end of the Galactic bar. For the
mean extinction of AV = 12.6 ± 0.5 mag (AK = 1.5 ± 0.1 mag), the color–magnitude diagram of probable cluster
members is well fit by isochrones in the age range 18–24 Myr. The estimated cluster mass is ∼20,000 M . With the
most massive original cluster stars likely deceased, no strong radio emission is detected in this vicinity. As such,
this red supergiant (RSG) cluster is representative of adolescent massive Galactic clusters that lie hidden behind
many magnitudes of dust obscuration. This cluster joins two similar RSG clusters as residents of the volatile region
where the end of our Galaxy’s bar joins the base of the Scutum-Crux spiral arm, suggesting a recent episode of
widespread massive star formation there.
Key words: Galaxy: stellar content – infrared: stars – open clusters and associations: general
(Stephenson 1990; Davies et al. 2007), which has 26 cluster
RSGs. The number of massive RSGs in a cluster places strong
limits on cluster age (massive stars have not all disappeared yet),
cluster mass (enough initial mass to produce large numbers of
high-mass stars), and the duration of the star formation burst
that produced the cluster, in that the stars must have formed in
a relatively short period to have reached the same evolved state
simultaneously. Figer et al. (2006) and Davies et al. (2007) use
an evolved initial mass function (IMF) to derive initial cluster
masses of 20,000–40,000 M for RSGC1 and RSGC2.
In this paper, we describe the discovery of another RSG
cluster using methods similar to those of Figer et al. (2006).
The cluster was discovered serendipitously as a conspicuous
grouping of bright red stars while perusing three-band midinfrared (mid-IR) color mosaics from GLIMPSE. At = 29.◦ 223
and b = −0.◦ 207, the cluster lies in roughly the same direction
as the RSG clusters reported in Figer et al. (2006) and Davies
et al. (2007). We use archive data from Spitzer, 2MASS, and
the GRS, along with new near-IR spectroscopy of nine probable
members to constrain the stellar content, age, and distance of
this young and massive Galactic cluster.
1. INTRODUCTION
The Spitzer Space Telescope (SST) and the Galactic
Legacy Infrared Mid-Plane Survey Extraordinaire (GLIMPSE;
Benjamin et al. 2003) have opened new windows into many
aspects of star formation and Galactic structure. The GLIMPSE
mid-IR survey and many complementary surveys such as MIPSGAL (Carey et al. 2005) at 24 μm, Two Micron All Sky Survey
(2MASS; Skrutskie et al. 2006), and the Boston University Five
College Radio Astronomy Observatory (BU-FCRAO) Galactic Ring Survey (GRS) in 13 CO (Jackson et al. 2006) together
constitute powerful new probes at 1–30 arcsec resolutions of
previously obscured components of the Milky Way. These overlapping surveys have produced many new discoveries including
a new globular cluster (Kobulnicky et al. 2005) and new young
open clusters (Mercer et al. 2005; Figer et al. 2006; Davies et al.
2007). A more complete census of star clusters enables a better
understanding of the Galaxy’s formation and structure, its star
formation history, and its current rate of star formation.
Massive stars are rare and short-lived. A typical 9 M star
will spend about 26 Myr on the main sequence (MS), while
a 15 M star has a MS lifetime of around 13 Myr (Meynet
& Maeder 2000). After leaving the MS these stars begin their
helium burning red supergiant (RSG) phase. The He-burning
lifetime is about 4 Myr for 9 M , 2.5 Myr for 15 M , and
even shorter for higher mass stars (Meynet & Maeder 2000).
This is a very small window for observing these evolved stars
and the reason why there are relatively few of them known in
the Galaxy. Bertelli et al. (1994) discovered five RSGs in NGC
7419, and more recently Figer et al. (2006) discovered 14 RSGs
and a yellow (G-type) supergiant in single Galactic cluster.
The current record holder for the most number of RSGs in a
single Galactic cluster is Stephenson 2, also known as RSGC2
2. DATA SETS EMPLOYED
2.1. Infrared Archival Data
Our discovery and analysis make use of archival data from
the SST GLIMPSE I/II and MIPSGAL I legacy surveys. The
GLIMPSE (Benjamin et al. 2003) project was conducted using
the Infrared Array Camera (IRAC; Fazio et al. 2004) wavebands
centered near 3.6, 4.5, 5.8, and 8.0 μm.3 With 4 s of integration
time per target, 5σ sensitivities of 0.2, 0.2, 0.4, and 0.4 mJy
3 GLIMPSE and MIPSGAL data are available from the Spitzer Science
Center Web site http://ssc.spitzer.caltech.edu/
4824
No. 6, 2009
THE DISCOVERY OF A MASSIVE CLUSTER OF RSGs WITH GLIMPSE
4825
-0.050
-0.100
-0.150
Galactic latitude
10
15
12
14
6
7
-0.200
11
5
16
1
3
4
2
9 13 8
-0.250
-0.300
10 pc
-0.350
29.400
29.350
29.300
29.250
Galactic longitude
29.200
29.150
29.100
Figure 1. Three-color IRAC and MIPS image of the cluster region with 4.5 μm in blue, 8.0 μm in green, and 24 μm in red. Numbered circles 1–9 denote cluster stars
with follow-up near-IR spectroscopy, while 10–16 are candidate cluster0 stars near the cluster core. The bar at lower right shows a linear scale of 10 pc at the adopted
cluster distance of 7.0 kpc.
were achieved for each band, respectively, on point sources
with angular resolutions of ∼2 . In total, the GLIMPSE I
implementation covers | | = 10◦ –65◦ and | b |< 1◦ . Each
image was calibrated by the Spitzer Science Center and further
processed by the GLIMPSE team pipeline to produce pointsource catalogs (PSCs) and 3◦ ×2◦ mosaics. The SST MIPSGAL
survey (Carey et al. 2005) covers the same region of sky as
GLIMPSE but uses the Multiband Infrared Photometer for
Spitzer (MIPS; Rieke et al. 2004) bandpasses at 24, 70, and
160 μm. The 24 μm mosaics have a resolution of 6 and a 5σ
sensitivity of 1.7 mJy (Carey et al. 2005).
Figure 1 shows a three-color image of the star cluster field,
with Spitzer IRAC [4.5] in blue, IRAC [8.0] in green, and
MIPS [24] in red. Blue predominantly highlights the stellar photospheres. Green shows diffuse emission from warm molecular
cloud interfaces, i.e., polycyclic aromatic hydrocarbon (PAH)
emission from photodissociation regions excited by UV photons. Red preferentially reveals warm dust. The scale bar in the
lower right shows a linear scale of 10 pc at a distance of 7.0 kpc.
The grouping of bright stars near the center of the image sug-
gests a physical association. They also lie near a local minimum
in the diffuse emission, possibly a cavity sculpted by the winds
and energetic photons from young massive stars. Infrared dark
clouds (IRDCs) and dark filaments appear superimposed on the
bright diffuse background at many locations across this field, indicating the presence of opaque, cold clouds in the foreground.
A handful of very red and bright sources, often coinciding with
dark clouds, surround the cluster. The color and placement of
these objects suggest that they are young stellar objects (YSOs)
associated with the IR emission.
We also use photometry from the University of Massachusetts
2MASS (Skrutskie et al. 2006), which observed the sky in the
near-IR for J (1.25 μm), H (1.65 μm), and Ks (2.16 μm) bands.
The project attained magnitude limits of J = 15.9, H = 15.0,
and Ks = 14.3, with a signal-to-noise ratio of 10.
Figure 2 shows a color composite image of the cluster with
the 2MASS J band in blue, the H band in green, and the
Ks band in red. The brightest cluster stars are conspicuous
as a grouping of yellow-red stars near the center of the
image.
4826
ALEXANDER ET AL.
Vol. 137
-0.140
-0.160
Galactic latitude
-0.180
-0.200
-0.220
-0.240
-0.260
-0.280
29.300
29.280
29.260
29.240
29.220
29.200
29.180
29.160
29.140
Galactic longitude
Figure 2. Enlarged three-color image of the cluster field with 2MASS J band in blue, H band in green, and Ks band in red.
Table 1
Cluster Members and Candidates
Star No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Statusa
GLIMPSE ID
R.A.(J2000)
Decl.(J2000)
J
H
Ks
[3.6]b
[4.5]b
[5.8]
[8.0]
S.T.
M
F
M
M
M
M
M
M
M
C
C
C
C
F
C
C
G029.1974 − 00.1975
G029.1973 − 00.1995
G029.2206 − 00.2037
G029.2139 − 00.2087
G029.2215 − 00.2075
G029.2380 − 00.1999
G029.2351 − 00.2059
G029.2290 − 00.2147
G029.2421 − 00.2154
G029.2076 − 00.1790
G029.2121 − 00.1935
G029.2286 − 00.1846
G029.2317 − 00.2137
G029.2385 − 00.1911
G029.2386 − 00.1854
G029.2480 − 00.2166
18:45:19.38
18:45:19.81
18:45:23.26
18:45:23.59
18:45:24.17
18:45:24.34
18:45:25.31
18:45:26.52
18:45:28.12
18:45:16.55
18:45:20.13
18:45:20.03
18:45:26.60
18:45:22.53
18:45:21.32
18:45:29.00
−03:24:48.1
−03:24:51.6
−03:23:44.0
−03:24:13.8
−03:23:47.2
−03:22:42.0
−03:23:01.0
−03:23:35.2
−03:22:54.5
−03:23:44.6
−03:23:54.1
−03:22:46.8
−03:23:24.7
−03:22:26.4
−03:22:16.2
−03:22:37.5
9.06
8.37
8.51
8.55
9.12
8.54
8.42
8.53
9.34
9.65
11.69
9.53
11.26
8.40
10.16
9.51
6.97
7.13
6.52
6.54
7.10
6.43
6.34
6.62
7.26
7.43
8.80
7.29
8.78
7.09
8.39
7.08
6.04
6.50
5.51
5.58
6.20
5.35
5.31
5.75
6.29
6.43
7.39
6.30
7.46
6.47
7.49
5.89
...
...
...
...
...
...
...
...
...
...
...
...
...
...
7.06
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
5.24
5.70
4.45
4.73
5.22
4.19
4.08
4.92
5.49
5.57
5.71
5.46
6.01
6.12
6.66
4.79
5.20
5.75
4.20
4.50
4.84
4.14
4.04
4.63
5.44
5.52
5.54
5.43
5.81
6.07
6.69
4.61
K5 I
A0 I
M1 I
M0 I
K4 I
M3 I
M2 I
K5 I
K4 I
...
...
...
...
...
...
...
Notes.
a Membership status. M denotes a member, F a foreground object, and C a cluster candidate.
b Ellipses indicate no listing in the GLIMPSE PSC owing to saturation.
Nine stars (numbered 1–9 in Figure 1) were chosen for
spectroscopic observation based on their location (clustering)
and brightness in the GLIMPSE/MIPSGAL mosaics. Table 1
lists these nine stars along with their equatorial coordinates,
2MASS JHK and GLIMPSE IRAC photometry. Most stars are
saturated in the IRAC bands.
2.2. Near-IR Spectroscopy
Near-IR spectroscopy was conducted using the Mimir instrument (Clemens et al. 2007) on the Perkins 1.8 m telescope, located outside Flagstaff, Arizona, on the UT nights
of 2008 May 21, June 25–28, and September 23. Mimir is a
No. 6, 2009
THE DISCOVERY OF A MASSIVE CLUSTER OF RSGs WITH GLIMPSE
cryogenic reimager with spectroscopic capability, delivering up
to a 10 × 10 arcmin field of view to a 1024 × 1024 InSb ALADDIN III array detector at 0.6 arcsec pixel−1 . Read noise is
typically 17–18 e− rms.
All spectroscopy used the JHK-grism (120 l mm−1 , 30◦
blaze angle, CaF2 substrate with 29◦ wedge) in conjunction
with one of three long-pass (LP) filters and the wide-field, F5,
camera optics to yield spectra with resolutions of 430–780,
for J band through K band, respectively. All May and some
June observations used the 1.16 μm LP filter to yield spectra
unaliased to 2.27 μm, while the remaining June observations
use the 1.90 μm LP to form spectra to 2.5 μm. September
observations used a 1.4 μm LP to simultaneously observe the
full H and K bands.
The slit used has projected sizes of 1.2 arcsec (2 pixels) by
5 arcmin and was oriented north–south. Spectra were typically
exposed for 15–45 s then moved along the spatial direction and
another spectrum obtained. This A-B cycle was repeated for
five or six pairs. The dome was rotated to cover the telescope
beam, and exposures were taken with continuum lamps on and
off in order to produce spectral flat fields. Spectral images of an
argon lamp positioned over the slit were obtained for wavelength
calibration. For calibration of the wavelength dependence of
atmospheric transmission, spectroscopy of bright A0V stars was
performed using all of the same steps, for stars chosen to be
within 15 degrees and 0.1 air masses of the target stars.
Data reduction used custom IDL-based programs that included measuring and correcting the nonlinear response of the
InSb photovoltaic pixels, using the fourth-order method described in Clemens et al. (2007), in addition to the usual dark
and flat-field corrections. Wavelength calibration utilized the argon spectra to ascertain the two-dimensional spatial-dispersion
coefficients necessary to remap the spectral images into uniformly dispersed versions. Typical dispersion solution scatter
was at or under 0.1 pixels (0.14 nm) for all argon spectra. Target
spectral images were flattened, interpolated, and remapped to
20 subpixels for each raw pixel, then rebinned to twice the original dispersion to achieve a fixed channel spacing and common
reference wavelength.
Spectra were modeled on the images as sloped lines with
“bow tie” spatially broadened extrema, due to the grism action.
Spectra were extracted from the modeled locations by fitting
locally sloped backgrounds at each spectral channel outside the
spatial extent covered by the spectra, then integrating the signal
above the background over the 2–3 rows showing significant
emission. Noise was estimated for each extracted spectral
channel based on Gaussian read noise and Poisson background
and signal contributions. This somewhat overestimates the noise
for these 2:1 synthetic channels. Relative scaling factors for each
of the 10–12 spectra for a target were found across either the Hor K-band. For each spectral channel, the 10–12 scaled values
were tested to remove bad pixels or cosmic rays, using median
filtering. The surviving data were averaged using weighting by
inverse noise squared.
Telluric spectra were developed using the A0V standard
star extracted spectra and the SpeXTool program XTelcor
by Cushing et al. (2004) and used to correct the target
star spectra. The final spectra were flattened by identifying
wavelengths free of lines, fitting baselines, and dividing by
the baseline functions to yield relative intensity spectra. The
resulting spectra, normalized by their error-propagated noise
spectra, showed signal-to-noise ratios of 200–450.
4827
Figure 3. Spectra for stars S1–S4. The wavelengths cover approximate H- and
K-band ranges. Our observations are the solid black lines, while smoothed atlas
spectra are shown as dotted lines. The spectral types for the atlas stars are given,
and the most prominent spectral features are labeled.
2.3. 13 CO Molecular Line Data
As a means of assessing the molecular environment of
the cluster, we used data cubes from the BU-FCRAO GRS4
(Jackson et al. 2006), which mapped the 13 CO J = 1 → 0
molecular line in the first Galactic quadrant from l = 18◦ to
55.◦ 7 and | b | 1◦ . The FCRAO 14 m telescope has a beam
size of 46 FWHM; the spectra were smoothed to a velocity
resolution of 0.21 km s−1 , and cover a local standard of rest
(LSR) velocity range from −5 to 135 km s−1 . The rms channel
sensitivity of the GRS is 0.4 K TA∗ .
3. ANALYSIS
3.1. Near-IR Spectra
Near-IR spectra of the first nine potential cluster members
are displayed in Figures 3 and 4. The solid lines show portions
of the normalized H-band and K-band spectra for each star,
shifted by arbitrary offsets for clarity. The dashed lines show
H-band spectra from Meyer et al. (1998) and K-band spectra
from Wallace & Hinkle (1997)5 for the nearest available spectral
type. The spectra of Meyer et al. (1998) and Wallace & Hinkle
(1997) have a resolution R = 3000, which we smoothed by
a factor of 10 to match the resolutions of our data. Eight
of the nine stars (S1 and S3–S9) show deep CO band head
features near 2.3 μm, characteristic of cool, evolved stars.
These spectra are most consistent with late-type giants or
supergiants.
Star S2, unlike the rest, is best matched with an early A or
late B supergiant spectrum. This determination is based on the
presence of strong Brackett series features in the H-band and
Brackett γ near 2.17 μm. The H-band Brackett lines of S2 were
only about half as wide as the same lines in the A0V comparison
stars, confirming its low-gravity, supergiant status. The lack of
4
GRS data can be obtained from the Boston University Web site
http://www.bu.edu/galacticring/.
5 The data were obtained from the NOAO Digital Library
http://www.lsstmail.org/dpp/library.html
4828
ALEXANDER ET AL.
Figure 4. Spectra for stars S5–S9, as in Figure 3.
Vol. 137
Figure 6. 2MASS J − H vs. H −Ks color–color diagram of the field in Figure 1.
The small dots designate every point source in Figure 1 with accurate (σ < 0.3
mag) JHKs photometry. The filled diamonds designate the nine stars with nearIR spectra. The filled circles denote other stars near the cluster core with Ks < 8.
The open triangles mark the fiducial unextincted supergiant sequence. The solid
line shows a reddening vector for AV = 12 mag.
show systematically lower EW than supergiants for a given
spectral type. The strong correlation between EW and spectral
type, along with the systematic differences between giants and
supergiants, shows that even at relatively low spectral resolution,
our IR spectra can be used to discern spectral and luminosity
classes. Typical values for the EW of cluster members are 45–
65 Å, which we used to assign the spectral types according to
the best-fit line; these are shown as solid diamonds in Figure 5.
We have estimated our uncertainty on the EW to be ±5 Å and
given the slope of the least-squares fit, this corresponds to ±2
subtypes. Table 1 lists these adopted spectral types.
Figure 5. Equivalent width of the CO band head vs. spectral type. The open
triangles show supergiants and open diamonds show giants from the stellar
spectral atlases of Meyer et al. (1998) and Wallace & Hinkle (1997). The solid
and dashed lines are least-squares fits to the supergiant and giant sequence,
respectively. The filled diamonds are our sources, placed along the best-fitting
line for supergiants. The cross represents our uncertainties of ±5 Å and ±2
subtypes for each star.
molecular absorption features (e.g., CO) is another indication
that S2 is an early-type star.
Following the method outlined by Figer et al. (2006) (see
their Figure 8), we measured the equivalent width (EW) of the
first 12 CO overtone feature at 2.30 μm for each star (except S2
as it shows no CO absorption). EW values were found by setting
the continuum level to be the average value of the normalized
flux between 2.28 and 2.30 μm and integrating the absorption
between 2.30 and the pseudocontinuum longward of 2.32 μm.
The same method was used for the observed stars as well as
stars with known spectral types from Wallace & Hinkle (1997).
Figure 5 shows the correlation between EW and spectral type
for the atlas stars of supergiant (open triangles) and giant (open
diamonds) luminosity classes. The solid line is a least-squares fit
to 14 G3–M6 supergiants, and the dashed line is a least-squares
fit to eight G7–M4 giants. This linear dependence extends
all the way down to G3 spectral classes, however, we only
show spectral types K0–M7 in Figure 5 for clarity. The giants
3.2. IR Photometry
Figure 6 is a 2MASS J − H versus H − Ks color–color
diagram showing the nine stars with spectral types (filled
diamonds), bright (Ks < 8) stars within ∼4 arcmin of the cluster
center (filled circles), and field stars in Figure 1 (dots). The open
triangles with labels denote standard JHKs colors for supergiant
stars.6 The dotted line is the reddening vector using the Cardelli
et al. (1989) extinction law, and the solid line is an empirical
reddening vector based on the slope of the locus of points. This
was done to attempt to correct for the deviation in the data from
the Cardelli et al. (1989) reddening. The difference between
the slopes corresponds to Δ(J − H ) = 0.25 over 12 mag of
visual extinction. Since J − H offset is small compared to our
estimated error, we still adopt the Cardelli et al. (1989) reddening
law for our data analysis. Eight of the nine stars with near-IR
spectral measurements occupy a limited region of the color–
color diagram—J − H 2.0, H − Ks 1.0—consistent
with similarly reddened late-type supergiants. S2 has a much
different color than the rest of the stars, suggesting an earlier
spectral type. Of the other bright stars without spectral types,
S14 is similar to S2 and less red than most cluster members,
while S11, S13, and S16 are more red, suggesting that they may
6
The photometric data are in the 2MASS JHKs color system, so we used
transformations from Carpenter (2001) and Bessell & Brett (1988) to convert
the data from Whittet & van Breda (1980), Elias et al. (1985), and Koornneef
(1983) to the 2MASS filter system.
No. 6, 2009
THE DISCOVERY OF A MASSIVE CLUSTER OF RSGs WITH GLIMPSE
Figure 7. Enlarged 2MASS color–color diagram as in Figure 6, with the fiducial
supergiant sequence reddened by AV = 12. A reddening vector for AV = 2
is shown. The reddened M-type supergiant standards match the colors of the
cluster stars. S2, an A0I, is less red than expected and is probably a foreground
object.
4829
Figure 9. 13 CO spectrum averaged over the area shown in Figure 1, showing
at least five 13 CO peaks along this sightline. Numbers above each peak denote
the near and far heliocentric kinematic distances in kpc based on the Clemens
(1985) rotation curve, while numbers in parentheses indicate the corresponding
Galactocentric distance. We identify the cluster with the molecular feature near
95 km s−1 (see the text).
1–2 mag. This indicates that S2 (and S14) are likely foreground
objects.
3.3. Molecular Environment and a Kinematic Distance
Figure 8. Color–magnitude diagram of stars in the field of Figure 1, with
labeling as in Figure 6. Reddened (Cardelli et al. 1989) isochrones with ages of
14 Myr (solid), 18 Myr (dotted), and 24 Myr (dashed) at a distance of 7.0 kpc
are shown. M-type supergiants match many of the cluster stars, while S2 is
less red and more than 2.5 mag brighter than a comparable A supergiant star
(off figure) at the same distance. The best-fit age is somewhere between 18 and
24 Myr. Note the presence of several stars (S10, S12, S15) with photometry but
lacking spectroscopy (filled circles) that have similar colors and magnitudes as
confirmed by M supergiants, suggesting the presence of several more supergiant
members. S14 was shifted up by 0.1 mag for clarity.
be more heavily extincted. Stars S10 and S12 have similar colors
to the majority of supergiants, suggesting that they may also be
RSG cluster members. Using the reddening law of Cardelli
et al. (1989), we converted the 2MASS colors, J − Ks and
H − Ks , into an average extinction. We find a mean extinction
of AV = 12.6 ± 0.5 (AKs = 1.5 ± 0.1).
Figure 7 is an enlarged version of Figure 6, with the fiducial supergiant colors reddened by AV = 12 using the empirical relation mentioned previously. Most of the stars show
colors matching K through M supergiants, with some differential extinction among cluster members. However, S2 does
not match the expected A0I supergiant at H − Ks = 0.77
and J − Hs = 2.1. S2 appears to be less reddened by
We used the GRS 13 CO data cubes to survey the molecular
environment near this RSG cluster. Figure 9 shows a plot of
the 13 CO brightness temperature versus LSR velocity averaged
across the field of view pictured in Figure 1. This figure shows
two strong 13 CO peaks at LSR velocities of roughly 79 and
95 km s−1 and several smaller peaks at 8, 47, 64, and
101 km s−1 . The abundance of 13 CO components shows that
this is a complex sightline with many molecular complexes
at various distances. Numbers above each peak indicate the
corresponding near and far heliocentric kinematic distances of
dkin = 5.0/9.8 kpc and 6.1/8.7 kpc for the two major peaks and
0.7/14.1 kpc, 3.2/11.6 kpc, 4.2/10.6 kpc, and ∼7.4 kpc7 for the
lesser peaks, respectively. The numbers in parentheses above
each peak indicate the Galactocentric distance for the given
velocity. These distances are derived from the Galactic rotation
curve of Clemens (1985), assuming R0 = 8.5 kpc, θ0 = 220
km s−1 . The more recent rotation curve of Levine et al. (2008)
yields very similar distances for the same R0 and θ0 . Adopting
R0 = 7.6 kpc instead (Eisenhauer et al. 2005) would reduce
all distances by about 10%. We estimate the kinematic distance
uncertainty at ±15% based on the width of the 13 CO features.
Maps of integrated intensity were created for each of the
five spectral features and compared to the GLIMPSE images.
Four of the features showed no correlation between the CO and
GLIMPSE images. For the fifth feature, with a peak at 95 km s−1 ,
a strong correlation of the CO distribution with the GLIMPSE
8 μm distribution was seen, as was a distinct anticorrelation
with the location of the stellar cluster.
In Figure 10, we overlay a map of the GRS 13 CO brightness
temperature at 95 km s−1 , averaged between 86 and 101 km
s−1 , on the GLIMPSE images. The GLIMPSE [4.5] and [8.0]
bands are shown in blue and green, respectively, while the 13 CO
is shown in red, along with white contours. This velocity range
exhibits a striking absence of molecular gas at the location of
7
This is the unique tangent-point distance corresponding to 103 km s−1 .
4830
ALEXANDER ET AL.
Vol. 137
-0.050
-0.100
-0.150
Galactic latitude
10
15
12
14
6
7
-0.200
11
5
16
1
3
4
2
9 13 8
-0.250
-0.300
10 pc
-0.350
29.400
29.350
29.300
29.250
Galactic longitude
29.200
29.150
29.100
Figure 10. Three-color image of the cluster region with 4.5 μm in blue, 8.0 μm in green, and the GRS 13 CO data averaged over the velocity range 86–101 km s−1
(dkin 6.1 kpc/8.7 kpc) in red with 13 CO contours in white. Contours range from TA∗ = 0.1 K to TA∗ = 20.1 K in increments of 2.0 K. The cluster lies near the 13 CO
minimum, suggestive of a cavity blown by the stellar winds of the massive cluster members.
the cluster while the surrounding regions are 13 CO-bright. This
13
CO minimum coincides with the 8.0 μm hole discussed earlier
in connection with Figure 1. We interpret this feature to be a
hole in the interstellar medium (ISM) formed by the winds of
massive stars that have already evolved over the lifetime of the
cluster. This velocity range, which peaks at 94 km s−1 , has a near
kinematic distance of 6.1 kpc and a far kinematic distance of
8.7 kpc. Distinguishing between these two possibilities requires
an independent distance measurement.
3.4. A Spectrophotometric Cluster Distance
Spectroscopic parallax is commonly used to find distances
when the relation between spectral type and absolute magnitude
is well known and has small dispersion (e.g., for MS stars).
Red supergiants, however, exhibit a large range of luminosities,
spanning several magnitudes at any given spectral type. This
is seen in the evolutionary tracks from Marigo et al. (2008),
which are nearly vertical at the coolest temperatures. The large
dispersion of absolute magnitudes among RSGs in Levesque
et al. (2005) supports this empirically. In order to estimate a
spectrophotometric distance for the cluster, we calculated the
mean absolute V magnitude and standard deviation for RSG
spectral types K3–M4.5 (representative of our sample) from
Levesque et al. (2005). By adopting a mean intrinsic V − K
color of 4.2 (Levesque et al. 2005), we derive a mean absolute
K magnitude for RSGs of MK = −10.0 ± 0.9. Given our mean
observed K magnitude of 5.75 and extinction AK = 1.5 for our
sample (excluding the possible foreground star S2), we estimate
a distance of 7.0+3.7
−2.4 kpc.
Since the spectrophotometric distance is highly uncertain,
we can only broadly constrain the distance of the cluster using spectroscopic parallax to the broad range 4–11 kpc, which
encompasses both kinematic distances. For the purposes of determining a cluster age below, we will adopt the spectrophotometric distance of 7.0 kpc.
3.5. Cluster Age
Figure 8 shows a Ks versus H −Ks color–magnitude diagram,
along with the fiducial supergiant sequence, at a distance of
No. 6, 2009
THE DISCOVERY OF A MASSIVE CLUSTER OF RSGs WITH GLIMPSE
4831
Table 2
Known RSG Clusters
Name
This cluster
RSGC1a
RSGC2b
(◦ )
b
(◦ )
R.A.(J2000)
Decl.(J2000)
N RSGs
Age
(Myr)
D
(kpc)
29.2
25.2
26.2
−0.20
−0.15
−0.06
18:45:20
18:37:58
18:39:20
−03:24:43
−06:52:53
−06:01:41
8–14
14
26
18–24
12 ± 2
17 ± 3
7.0
5.8,6.5c
5.8
AV
(mag)
∼12
24
13
Mass
(M )
RGal
(kpc)
∼20,000
30,000 − 40,000
40,000
3.9
3.5
3.5
Notes.
a Figer et al. (2006).
b Davies et al. (2007).
c Nakashima & Deguchi (2006).
7.0 kpc and reddened by AV = 12. Symbols are the same as
Figure 6. The solid, dotted, and dashed curves are 14, 18, and
24 Myr isochrones8 from Marigo et al. (2008). They have been
reddened by AV = 12 and placed at a distance of 7.0 kpc. The
majority of the cluster stars are consistent with being reddened
supergiants at 7.0 kpc for ages between 14 and 24 Myr. Note
that two objects with photometry but lacking spectroscopy, S10
and S12, fall in the same region as spectroscopically confirmed
RSGs, bringing the total number of probable red supergiants in
the cluster to 10. Another object, S15, closely matches the G0I
fiducial point. There are five other candidates, S2, S11, S13, S14,
and S16 that lie very close to the center of the cluster spatially,
but do not occupy the same color-magnitude space. The A0 I star,
S2, is much less red and nearly 2 mag brighter than it should be
if it were a cluster member. As stated previously we consider S2
to be a foreground star, not associated with the cluster. Based on
the same criteria (but lacking spectroscopy) S14 is also likely
to be a foreground object. Stars S11, S13, and S16 may be
background objects or possibly cluster members enshrouded
by additional intracluster or circumstellar dust. Massey et al.
(2005) show that supergiants may be surrounded by up to 5
visual magnitudes of circumstellar extinction. Thus, we consider
stars S11, S13, and S16 as candidate cluster supergiants,
given their brightness and proximity to the cluster core. From
Figure 8 we are then able to conclude that the cluster contains
eight spectroscopically confirmed supergiants as well as six
additional cluster candidates.
The most massive stars remaining in each isochrone are
18.6 M , 14.9 M , and 12.6 M corresponding to ages of
14 Myr, 18 Myr, and 24 Myr. These stars are the most luminous
objects in the cluster in the near-IR, so the excellent agreement
between the tips isochrones and the cluster supergiants indicates
that the data are consistent with an age range of 18–24 Myr.
Examination of the 20 cm radio continuum map of this region
from the NRAO VLA Sky Survey (Condon et al. 1998) shows
no sources of emission at the location of the cluster, consistent
with an older age and lack of ionizing photons from the most
massive stars.
3.6. Cluster Stellar Mass
The eight confirmed and 3–6 additional candidate RSGs
in the cluster imply a large total stellar mass, and, given the
short relative lifetime of the RSG phase, a short star formation
timescale. Davies et al. (2008) compute Monte Carlo population
synthesis models with a Salpeter IMF (Salpeter 1955) and
Geneva isochrones (Meynet & Maeder 2000) to estimate the
number of RSGs as a function of age for a range of initial
8
Isochorones obtained from http://stev.oapd.inaf.it/cmd
cluster masses. Based on their Figure 8, a cluster containing
10 RSGs at an age of 18–24 Myr has a initial mass of
∼20,000 M . In this cluster, all stars except the most luminous
supergiants are faint due to its distance and reddening, which
means the underlying cluster is below the detection threshold for
2MASS.
4. DISCUSSION AND CONCLUSIONS
This newly discovered cluster resembles two other RSG
clusters in terms of stellar mass, age, the high number of red
supergiants, and placement in the Galaxy. Table 2 summarizes
the locations and derived parameters for the clusters RSGC1
(Figer et al. 2006; Davies et al. 2008), RSGC2 ≡ Stephenson 2
(Stephenson 1990; Davies et al. 2007), and the present cluster.
We have listed the distances for each cluster based on published
data and computed the Galactocentric distance, RGal , based on a
solar Galactocentric distance of 7.6 kpc (Eisenhauer et al. 2005).
Although the uncertainties of the spectrophotometric distances
are several kiloparsecs, the resulting locations at = 25◦ –29◦
and Rgal = 3.5–3.7 kpc indicate that these clusters may lie
within a few hundred parsecs of each other. Davies et al. (2007)
suggest that this is where the Scutum-Crux spiral arm meets the
Galactic bulge and is within a stellar ring proposed by Bertelli
et al. (1995). This location may also lie near the northern end
of the Galactic bar, the parameters of which are still uncertain.
There is evidence for at least two nonaxisymmetric structures
in the inner Galaxy, a vertically thick triaxial bulge/bar of halflength 3.5 kpc and angle 20–35 deg (Bissantz & Gerhard 2002;
Gerhard 2001; Babusiaux & Gilmore 2005) and a vertically
thinner “Long Bar” of half-length 4.4 kpc and angle 40–45
deg (Hammersley et al. 2000; Benjamin et al. 2005). RSGC1,
RSGC2, and this new cluster lie near the end of the Long Bar,
perhaps where it meets the Scutum-Crux spiral arm and or the
Molecular Ring.
Figure 11 plots the locations of RSGC1 and RSGC2 (asterisk)
and the newly discovered cluster (triangle). The thick dashed
lines show the spiral arms as defined by Nakanishi & Sofue
(2006). The thin contours show the stellar bar as modeled by
Bissantz & Gerhard (2002) from 2MASS near-IR photometry,
while the thick solid curve shows the 4.4 kpc radius bar inferred
by Benjamin et al. (2005) from Section ST mid-IR photometry.
The location of all three RSG clusters near the end of the bar is
consistent with recent star formation, possibly induced by radial
gas flows along the bar. Numerical simulations show radial
flows resulting from barred potentials (Schwarz 1981; Combes
& Gerin 1985; Englmaier & Gerhard 1997; Piner et al. 1995;
Rodriguez-Fernandez & Combes 2008). Observational studies
of star formation in other galaxies also reveal enhanced activity
associated with bars, sometimes along the bars themselves and
4832
ALEXANDER ET AL.
Vol. 137
California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science
Foundation. This publication makes use of molecular line data
from the Boston University Five College Radio Observatory
Galactic Ring Survey (GRS), funded by the National Science
Foundation under grants AST-9800334, −0098562, −0100793,
−0228993, and −0507657. This research was conducted in
part using the Mimir instrument, jointly developed at Boston
University and Lowell Observatory and supported by NASA,
NSF, and the W. M. Keck Foundation. This work was partially supported by the NSF under grant AST-0607500 to Boston
University.
Note added in proof. As this paper went to press, we became
aware of the simultaneous discovery of this star cluster by Clark
et al. (2009).
REFERENCES
Figure 11. Schematic of face-on Milky Way showing the locations of RSGC1
and RSGC2 (asterisk) and the newly discovered cluster (triangle). The symbol
sizes represent the uncertainties in the clusters’ locations. The thick dashed
curves denote spiral arms as described by Nakanishi & Sofue (2006). The thin
contours show the triaxial/bulge bar as modeled by Bissantz & Gerhard (2002),
while the thick solid curve shows the thin “Long Bar” as derived by Benjamin
et al. (2005).
sometimes near the ends of bars where spiral arms or “spurs”
attach to the bars (Martin & Friedli 1997; Sheth et al. 2002).
In cases where star formation occurs along bars, it is displaced
toward the leading side of the stellar bar and peaks outside of
the dust and molecular lanes that run along the bar (Martin &
Friedli 1997; Sheth et al. 2002). For example, Downes et al.
(1996) present CO maps of the barred spiral galaxy NGC 1530
showing concentrations of molecular material near the bar ends
at ∼4 kpc from the nucleus. The accompanying optical image
appears to show enhanced star formation just outside this region
near the bar ends. Sheth et al. (2000) report that Hα emission
peaks near the ends of the ∼4 kpc stellar bar in NGC 5383.
The locations of the RSG clusters in Figure 11 near one end
of the Milky Way’s bar at a ∼4 kpc radius hint that a similar
phenomenon may be happening in our galaxy.
Taken together, the discovery of this new cluster raises
the total count of RSGs in the Galaxy and adds evidence
for a localized burst of recent massive star formation. The
accumulating data point toward widespread, high-mass star
formation in this vicinity over the last ∼20 Myr. With a stellar
mass of roughly 20,000 M and a population containing 8–14
red supergiants, this ranks among the more massive clusters in
the Galaxy.
We would like to thank Charles Kerton & Kim Arviddson
for assisting with the initial discovery of this cluster and for
software help. We acknowledge Bob Benjamin and Danny
Dale for helpful comments. We thank an anonymous referee
for a rapid and expert review. H.A.K. and M.J.A. were supported by NASA through grant NAG5-10770. This publication makes use of data products from the Two Micron All
Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/
Babusiaux, C., & Gilmore, G. 2005, MNRAS, 358, 1309
Benjamin, R. A., et al. 2003, PASP, 115, 953
Benjamin, R. A., et al. 2005, ApJ, 630, L149
Bertelli, G., Bressan, A., Chiosi, C., Fagotto, F., & Nasi, E. 1994, A&AS, 106,
275
Bertelli, G., Bressan, A., Chiosi, C., Ng, Y. K., & Ortolani, S. 1995, A&A, 301,
381
Bessell, M. S., & Brett, J. M. 1988, PASP, 100, 1134
Bissantz, N., & Gerhard, O. 2002, MNRAS, 330, 591
Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245
Carey, S. J., et al. 2005, BAAS, 37, 1252
Carpenter, J. M. 2001, AJ, 121, 2851
Clark, J. S., et al. 2009, arXiv:0903.1754
Clemens, D. P. 1985, ApJ, 295, 422
Clemens, D. P., Sarcia, D., Grabau, A., Tollestrup, E. V., Buie, M. W., Dunham,
E., & Taylor, B. 2007, PASP, 119, 1385
Combes, F., & Gerin, M. 1985, A&A, 150, 327
Condon, J. J., Cotton, W. D., Greisen, E. W., Yin, Q. F., Perley, R. A., Taylor,
G. B., & Broderick, J. J. 1998, AJ, 115, 1693
Cushing, M. C., Vacca, W. D., & Rayner, J. T. 2004, PASP, 116, 362
Davies, B., Figer, D. F., Kudritzki, R.-P., MacKenty, J., Najarro, F., & Herrero,
A. 2007, ApJ, 671, 781
Davies, B., Figer, D. F., Law, C. J., Kudritzki, R.-P., Najarro, F., Herrero, A., &
MacKenty, J. W. 2008, ApJ, 676, 1016
Downes, D., Reynaud, D., Solomon, P. M., & Radford, S. J. E. 1996, ApJ, 461,
186
Eisenhauer, F., et al. 2005, ApJ, 628, 246
Elias, J. H., Frogel, J. A., & Humphreys, R. M. 1985, ApJS, 57, 91
Englmaier, P., & Gerhard, O. 1997, MNRAS, 287, 57
Fazio, G. G., et al. 2004, ApJS, 154, 10
Figer, D. F., MacKenty, J. W., Robberto, M., Smith, K., Najarro, F., Kudritzki,
R. P., & Herrero, A. 2006, ApJ, 643, 1166
Gerhard, O. E. 2001, in ASP Conf. Ser. 230, Galaxy Disks and Disk Galaxies,
ed. J. G. Funes, S.J., & E. M. Corsini, (San Francisco, CA: ASP), 21
Hammersley, P. L., Garzón, F., Mahoney, T. J., López-Corredoira, M., & Torres,
M. A. P. 2000, MNRAS, 317, L45
Jackson, J. M., et al. 2006, ApJS, 163, 145
Kobulnicky, H. A., et al. 2005, AJ, 129, 239
Koornneef, J. 1983, A&A, 128, 84
Levesque, E. M., Massey, P., Olsen, K. A. G., Plez, B., Josselin, E., Maeder, A.,
& Meynet, G. 2005, ApJ, 628, 973
Levine, E. S., Heiles, C., & Blitz, L. 2008, ApJ, 679, 1288
Marigo, P., Girardi, L., Bressan, A., Groenewegen, M. A. T., Silva, L., &
Granato, G. L. 2008, A&A, 482, 883
Martin, P., & Friedli, D. 1997, A&A, 326, 449
Massey, P., Plez, B., Levesque, E. M., Olsen, K. A. G., Clayton, G. C., &
Josselin, E. 2005, ApJ, 634, 1286
Mercer, E. P., et al. 2005, ApJ, 635, 560
Meyer, M. R., Edwards, S., Hinkle, K. H., & Strom, S. E. 1998, ApJ, 508,
397
Meynet, G., & Maeder, A. 2000, A&A, 361, 101
Nakanishi, H., & Sofue, Y. 2006, PASJ, 58, 847
Nakashima, J. I., & Deguchi, S. 2006, ApJ, 647, L139
Piner, B. G., Stone, J. M., & Teuben, P. J. 1995, ApJ, 449, 508
No. 6, 2009
THE DISCOVERY OF A MASSIVE CLUSTER OF RSGs WITH GLIMPSE
Rieke, G. H., et al. 2004, ApJS, 154, 25
Rodriguez-Fernandez, N. J., & Combes, F. 2008, A&A, 489, 115
Salpeter, E. E. 1955, ApJ, 121, 161
Schwarz, M. P. 1981, ApJ, 247, 77
Sheth, K., Regan, M. W., Vogel, S. N., & Teuben, P. J. 2000, ApJ, 532,
221
4833
Sheth, K., Vogel, S. N., Regan, M. W., Teuben, P. J., Harris, A. I., & Thornley,
M. D. 2002, AJ, 124, 2581
Skrutskie, M. F., et al. 2006, AJ, 131, 1163
Stephenson, C. B. 1990, AJ, 99, 1867
Wallace, L., & Hinkle, K. 1997, ApJS, 111, 445
Whittet, D. C. B., & van Breda, I. G. 1980, MNRAS, 192, 467